unofficial linux-ck repo with cpu optimized packages is ONLINE

xander · 2016-10-29 16:55:09

Hello there. linux-ck-skylake-4.8.5 update actually slowed down system startup, DE (GNOME 3.22.1) loading and user login/logout time. 4.8.4, both -ARCH and -ck, works pretty good. Unfortunately I cannot say, on which pkgrel, -1 or -3, it firstly appeared. Also there are some messages on 4.8.5-3 related to 'watchdog' just before system reboots, but I can't exactly found it in systemd journals.
Any ideas? i5-6600K.

graysky · 2016-10-29 17:06:23

1) Boot into 4.8.4-1-ARCH
2) Post the output of `systemd-analyze critical-chain`
3) Boot into 4.8.5-3-ck
4) Post the output of `systemd-analyze critical-chain`

xander · 2016-10-29 17:22:27

Here it is.
4.8.4-1-ARCH:

The time after the unit is active or started is printed after the "@" character.
The time the unit takes to start is printed after the "+" character.

graphical.target @976ms
└─gdm.service @962ms +13ms
  └─systemd-user-sessions.service @960ms +2ms
    └─network.target @952ms
      └─NetworkManager.service @824ms +127ms
        └─dbus.service @773ms
          └─basic.target @761ms
            └─sockets.target @761ms
              └─org.cups.cupsd.socket @761ms
                └─sysinit.target @760ms
                  └─systemd-timesyncd.service @635ms +124ms
                    └─systemd-tmpfiles-setup.service @629ms +5ms
                      └─local-fs.target @629ms
                        └─boot-efi.mount @620ms +9ms
                          └─systemd-fsck@dev-disk-by\x2duuid-D021\x2d2A90.service @533ms +70ms
                            └─dev-disk-by\x2duuid-D021\x2d2A90.device @518ms

4.8.5-3-ck:

The time after the unit is active or started is printed after the "@" character.
The time the unit takes to start is printed after the "+" character.

graphical.target @1.007s
└─gdm.service @996ms +11ms
  └─systemd-user-sessions.service @994ms +1ms
    └─network.target @993ms
      └─NetworkManager.service @940ms +53ms
        └─dbus.service @906ms
          └─basic.target @881ms
            └─sockets.target @881ms
              └─dbus.socket @881ms
                └─sysinit.target @880ms
                  └─systemd-timesyncd.service @509ms +370ms
                    └─systemd-tmpfiles-setup.service @502ms +6ms
                      └─local-fs.target @502ms
                        └─home.mount @424ms +77ms
                          └─systemd-fsck@dev-disk-by\x2duuid-22a7dcc5\x2d0f4e\x2d449b\x2dbc4b\x2d2087feeb29e2.service @401ms +22ms
                            └─dev-disk-by\x2duuid-22a7dcc5\x2d0f4e\x2d449b\x2dbc4b\x2d2087feeb29e2.device @399ms

(feeling uncomfortable without [spoiler] BB)
The real slowdown feels much larger than it is on systemd stats, especially between the moment when GDM starts (both autologin and manual) and final desktop appearance.

graysky · 2016-10-29 17:58:26

You could post to ck's blog to let him know... but 976 ms is the same as 1,007 ms based on the critialchain output.

Jristz · 2016-10-29 21:53:49

Greysky im reading Con Kolivas (CK) has released MUQSS and as I underestand's a successor for BFS so you going to keep the patches bfq and bfs as is or move to bfq and muqss?

graysky · 2016-10-29 21:55:41

@Jristz - The ck patchset includes BFQ and MuQSS. BFS has been morphed into MuQSS so BFS is no more.

WFV · 2016-10-31 01:40:37

EDIT: see #2865

Last edited by WFV (2016-11-11 02:35:43)

xander · 2016-11-02 19:11:10

@graysky,
there is one new important commit in 4.8-ck branch, will you update PKGBUILD and maybe repo packages with this one in addition to two earlier?

graysky · 2016-11-02 19:13:00

@xander - Do you mean: https://github.com/ckolivas/linux/commi … 8e501e00ff

xander · 2016-11-02 19:17:48

Yes, exactly. Con made it after my comments about GDM/gnome-shell login slowdown in his blog. Despite the fact that it didn't helped me with login (and about an hour ago I've built several kernels with different Hzs and found that it's GNOME issues, not MuQSS), I can see some general speedups on manual-built linux-ck with this patch.

graysky · 2016-11-02 19:22:31

OK. I emailed CK asking about scope and stability. I don't see 4.8.7-rc1 yet so (I don't even see anything in the 4.8-stable queue), so if CK deems it safe and helpful, I'll bump including it. Thanks for the heads up.

graysky · 2016-11-02 23:10:42

Con emailed me back and in his opinion, 8a501e00 would only affect a niche group of users. I will apply it when 4.8.7 is released or if he rolls out a ck6.

xander · 2016-11-05 08:27:10

linux-ck6 with MuQSS 0.135 is out!

graysky · 2016-11-05 10:30:23

xander wrote:

linux-ck6 with MuQSS 0.135 is out!

https://aur.archlinux.org/cgit/aur.git/ … ad75001007

WFV · 2016-11-06 00:51:14

moved to CK's blog per:

graysky wrote:

@WFV - best to post that to ck's blog

Last edited by WFV (2016-11-11 02:36:32)

graysky · 2016-11-06 10:31:19

@WFV - best to post that to ck's blog

tpfkanep · 2016-11-08 05:03:29

Also reverted back to 4.8.6.1

Boot takes a lot longer. On a 32GB SSD, 4.8.6.1 systemd-analyze = under 3 secs. 4.8.6.2 seems to be stuck on fsck on boot or root (not too sure which) and the modules load. This results in a boot of 40+ seconds.

Jannis · 2016-11-11 16:18:18

Is there really a difference between the header packages of different architectures? I thought the packages only differed in compiler optimizations but that wouldn't affect the headers, would it? Or do the different packages actually have CPU-specific code changes? Just asking because I accidentally had the wrong header package installed but my third party kernel modules seemed to built and work fine, so that made me curious.

graysky · 2016-11-11 20:42:03

@Jannis - I believe so but I haven't played in a while. Might as well just keep the matched sets as I build them that way.

tpfkanep · 2016-11-16 09:02:56

I dunno if timer frequency affects boot time, but:
MuQSS with 100HZ Timer frequency = > 5 seconds boot.
MuQSS with 1000HZ Timer frequency = < 3 seconds boot.

4.8.8-1 kernel

Not related to the above: I have also noted that ejecting FAT usb-drives [eject /dev/sdx] takes longer in linux-ck when compared to the vanilla kernel. Any reason for this or is it something peculiar to my system?

plaxx · 2016-11-18 03:30:32

I have PGP signature errors on the haswell 4.8.8-1 packages:

error: linux-ck-haswell: signature from "graysky (used to sign repo-ck packages) <graysky@archlinux.us>" is invalid
:: File /var/cache/pacman/pkg/linux-ck-haswell-4.8.8-1-x86_64.pkg.tar.xz is corrupted (invalid or corrupted package (PGP signature)).
Do you want to delete it? [Y/n] n
error: linux-ck-haswell-headers: signature from "graysky (used to sign repo-ck packages) <graysky@archlinux.us>" is invalid
:: File /var/cache/pacman/pkg/linux-ck-haswell-headers-4.8.8-1-x86_64.pkg.tar.xz is corrupted (invalid or corrupted package (PGP signature)).
error: linux-ck-haswell: signature from "graysky (used to sign repo-ck packages) <graysky@archlinux.us>" is invalid
:: File /var/cache/pacman/pkg/linux-ck-haswell-4.8.8-1-x86_64.pkg.tar.xz is corrupted (invalid or corrupted package (PGP signature)).
Do you want to delete it? [Y/n] n
error: linux-ck-haswell-headers: signature from "graysky (used to sign repo-ck packages) <graysky@archlinux.us>" is invalid
:: File /var/cache/pacman/pkg/linux-ck-haswell-headers-4.8.8-1-x86_64.pkg.tar.xz is corrupted (invalid or corrupted package (PGP signature)).

Here are the hashes of what was downloaded:

$ sha1sum /var/cache/pacman/pkg/linux-ck-haswell*4.8.8-1-x86_64.pkg.tar.xz 
7aef2548404e33638786c1a909f7116f3de543c2  /var/cache/pacman/pkg/linux-ck-haswell-4.8.8-1-x86_64.pkg.tar.xz
0e45c5eb6af12932a5fd3c5ca8aa5171e83ef8b3  /var/cache/pacman/pkg/linux-ck-haswell-headers-4.8.8-1-x86_64.pkg.tar.xz

I hope this is the right place to report this. Thanks!

illis · 2016-11-18 22:31:13

I also got a PGP error with the 4.8.8-1 haswell package. Deleted it, and tried to re-download and now im getting a 'Maximum file size exeeded' error:

error: failed retrieving file 'linux-ck-haswell-4.8.8-1-x86_64.pkg.tar.xz' from repo-ck.com : Maximum file size exceeded

graysky · 2016-11-18 23:00:13

@plaxx and illis - Try now. Thanks for reporting.

arun_kpm · 2016-11-19 15:35:02

this is probably a stupid question,
If I compile a linux-ck kernel from AUR on a core2 machine, does it have a core2 specific optimizations or a generic ck kernel?
Thanks

Last edited by arun_kpm (2016-11-19 15:39:46)

graysky · 2016-11-19 15:47:29

arun_kpm wrote:

If I compile a linux-ck kernel from AUR on a core2 machine, does it have a core2 specific optimizations or a generic ck kernel?

Not without you manually editing the PKGBUILD to enable '_makenconfig=' and then selecting the optimization from the nconfig.

Arch Linux

#2851 2016-10-29 16:55:09

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2852 2016-10-29 17:06:23

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2853 2016-10-29 17:22:27

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2854 2016-10-29 17:58:26

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2855 2016-10-29 21:53:49

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2856 2016-10-29 21:55:41

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2857 2016-10-31 01:40:37

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2858 2016-11-02 19:11:10

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2859 2016-11-02 19:13:00

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2860 2016-11-02 19:17:48

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2861 2016-11-02 19:22:31

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2862 2016-11-02 23:10:42

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2863 2016-11-05 08:27:10

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2864 2016-11-05 10:30:23

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2865 2016-11-06 00:51:14

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2866 2016-11-06 10:31:19

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2867 2016-11-08 05:03:29

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2868 2016-11-11 16:18:18

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2869 2016-11-11 20:42:03

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2870 2016-11-16 09:02:56

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2871 2016-11-18 03:30:32

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2872 2016-11-18 22:31:13

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2873 2016-11-18 23:00:13

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2874 2016-11-19 15:35:02

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#2875 2016-11-19 15:47:29

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

Board footer